Another important point is the "twisted pair" requirement. The
question seems to have an implied EMI (Electro-Magnetic Interference
flavor.? On that assumption, I mention that not only is the radiated
power dependent on the frequency components of the current in the
wires, there is a spacial factor as well.?? High rate twists will
tend to cancel the low frequency radiation patterns if the current
"out" is balanced by the current "back".? I.e., the area of the loop
formed by the current loop of the wire spacing and that loop can
include the earth return if the current is essentially outbound from
the source and coming back in some other path than the 3 wires.?
Unless the 3 wires are a differential pair with a single wire
shield, you are pretty much assured of unbalance and higher radiated
energy.? Modeling the spatial factors in LTspice is not something I
would even begin to try.? RF modelling software that includes
3-dimensional object modelling along with electrical factors is
called for in this instance.
Regards,
Charles Patton
?
flpierson
wrote:
? ? "What I
need to do is take the FFT of the average current of
the three wires. How do I go about doing that?"
First,
define what you mean by "the average current of the
three wires."
I don't
even know what "average current" means in the context of
an FFT.? Usually an FFT operates on a waveform.? When
you average that waveform, you no longer have the
waveform; you have just a number.? You can't take an FFT
of a number.? Does "average" mean "filtered" or
"smoothed"?? Why bother doing that?? Can't you just take
the FFT of the current as it is?
What do
you mean by "the three wires"?? Do you want the FFT of
the current through each wire -- hence, three FFT
spectra?? Or do you want the FFT of something like the
net current of all three wires combined?
In
LTspice it's easy to get the FFT of any number of
signals..? Plot the ones you want on a waveform plot,
then right-click > View > FFT.? Here, you can
choose how many data samples, the time range over which
to evaluate the FFT, whether to do a wee bit of
smoothing by averaging adjacent time points, and whether
to apply a windowing filter.? The latter is unnecessary
if you've chosen the time span to include an exact
integer number of cycles.? The waveforms that were
plotted are highlighted in the top, but you can change
which one or ones you want.
Click
OK.? Now, if you had two or more waveforms selected,
you'll get another window that lets you choose again
exactly which ones to plot.? If you want all three,
just highlight them all.? Or you can enter an
expression here (such as I(R1)+I(R2)+I(R3)).
Click
OK.? Voila, there's your FFT spectrum or spectra.
Now,
if the thing you want is the FFT of the combined
current, figure out what that means to you.? Do you
want the common-mode current?? Differential mode??
Make an expression for what you want.? Then implement
it.? You could do that with a Bv source added to your
schematic (so that it exists as a separate signal that
can be probed), or you can enter the expression later
in the FFT process.
If you
are new to FFTs, make sure that you do all the things
necessary to get a good FFT.? Choose an appropriately
small Maximum Timestep.? Disable waveform compression
by adding ".options plotwinsize=0" to your schematic
(it's essential!).? Make sure to use an integral
number of cycles in the time interval passed to the
FFT.? If there are start-up transients, wait for those
to die out before starting the FFT.? Choose the total
time interval wisely, as it affects the FFT's
appearance -- one cycle is probably too little, 1000
cycles is probably too much.? Finally, LTspice's FFT
shows you tons of high-order harmonics that may have
no meaning, so ignore them, unless you are sure that
the waveforms have meaningful data up there at those
frequencies.? I guess the philosophy is it's better to
start with too much data and discard what you don't
want, than to start with too little and not know that
there was more.
